Rape Weight Detection Based On Image Processing

Rape is an important oil crop in China because of its high economic value and great development potential.Seed test is an important means in the process of rape breeding,and 1000-seed weight is an important parameter of rape yield estimation,which is of great significance to increase and yield of rape.The thousand-grain weight is generally divided into two steps: counting grains and weighing.The traditional1000-grain weight measurement is completed by manual counting and weighing,which is low in efficiency,time-consuming and low in accuracy.However,the image processing method based on machine vision can replace the process of manual processing,and effectively solve the above problems.Due to the high hardware cost and complex construction,the traditional system based on image processing has low portability in image acquisition.In this paper,smart phone are used to directly collect seed images of rape,which greatly improves mobile portability and reduces operation difficulty.Conventional 1000-seed weight determination based on image processing is easily affected by light intensity and noise,and the recognition and segmentation results of seed image are not ideal.In this paper,the following processes are used to optimize the identification,segmentation and counting process of rape seeds :(1)the image pretreatment method,the use of filtering algorithm and gray adjustment method,fully reduce the impact of noise in the image,enhance the observation characteristics of the image target area,reduce the subsequent identification and analysis difficulty.(2)Two methods based on limit corrosion and distance change and gray reconstruction are proposed for image segmentation of various particle target regions.(3)The area threshold algorithm was used to count the segmented seed regions.The method proposed in this paper solved the problem of incomplete particle adhesion and segmentation in the process of seed image processing of rape,and the data and results were verified based on Matlab image software.